A feature correlation-based fusion method for fingerprint and palmprint identification systems

The paper proposes an innovative approach for biometric systems design in which the identification is performed using a feature-level fusion method which discards the most correlated features. The accuracy improvement is ensured by the proposed fusion method and the hierarchical classification model which performs people recognition in 2 stages. This approach is suitable for biometric applications with different security requirements, such as the remote access control to medical databases in which the users have different authorization degrees and the misidentification costs are not the same for all the authorized users.